Ninja Copper: Scythe's 5th Year Celebration

Cooling
Viewing page 6 of 7 pages. Previous 1 2 3 4 5 6 7 Next

FURTHER INVESTIGATIONS

So flummoxed were we with the results that we felt compelled to carefully re-examine... everything, from the various heatsinks to every aspect of our testing procedure. This troubleshooting process took nearly a month to complete, partly because it could not be done continuously, and partly because the procedures were repeated many times to eliminate error. At the heart was the simple, somewhat incredulous question, "How can a brand new all-copper version of the Ninja be bettered by a beat-up old, aluminum sample?"

The first check was to run the NinjaCU test again, beginning with a completely fresh re-installation. And again and again. In all, it was installed and reinstalled four times. The performance never varied by more than 1°C.

Second check: Run the original Ninja test again. This time it tested about 1°C poorer at most fan speeds. This is probably beyond the resolving limits of our test system. It was uninstalled, reinstalled, and tested again. The -1°C results were repeated. OK, maybe this sample has been beaten up in the lab enough that it's lost a little something.

The above checks helped to confirm that the testing platform and our methodology were working as intended. Whew!

Third check: Obtain a new sample of the Ninja Ninja Plus Rev. B so that we could use its 478 mounting clips (very similar to the NinjaCU clips) with the 775 conversion hardware from the NinjaCU package. Much thanks to Vancouver retailer Anitec for quickly contributing a Ninja sample. (Note: It was only for this reason that the Ninja Plus Rev. B came to be in the photos on the preceding pages.) The results of this test were shocking. The Plus Rev. B performed nothing like the original or the Copper. The test was repeated several times, always with the same results.

Three Ninjas on socket 478 clips, retested: °C Rise
Reference 120mm Fan
12V
9V
7V
5V
NinjaCU
17
18
20
23
Original Ninja
15
16
19
21
Ninja Plus Rev. B
24
25
28
31

Fourth check: The flatness of each Ninja base was checked against a steel ruler. All three were flat enough that no gaps could be seen between the ruler edge and the base. The Ninja Plus Rev. B base seemed unusually smooth, perhaps too smooth. Thermal interface material is supposed to work best when there are gaps to fill...

ANALYSIS: SORTING IT OUT

After all that testing, can we make sense of it all? Yes. Sort of. Let's post one last table here, of the various Ninjas and comparables.

Scythe NinjaCU vs Competitors: °C Rise
Reference 120mm Fan
12V
9V
7V
5V
NinjaCU
17
18
20
23
Original Ninja (new test)
15
16
19
21
Ninja Plus Rev. B
24
25
28
31
Scythe Mugen
18
19
21
n/a
Thermalright Ultra-120
15
17
21
26
Thermalright Ultra-120 EX
12
14
17
24
Asus Triton 75
18
20
23
30

Conclusion #1: The new Ninja Copper is an excellent heatsink. It is bettered by our old original Ninja (which we believe is an exceptional sample), and by the Thermalright Ultra 120s at the moderate airflow represented by the Nexus 120 fan at 12V. With extremely low airflow, in essentially silent operation, the NinjaCU is bettered only by the old Ninja and matched by the Thermalright Ultra-120 Extreme. The Ninja's size and wide spacing between fins make the difference here. The supplied fan is an excellent match, running at a measured 18 dBA@1m even at 12V. It is the quietest fan we've seen supplied with any heatsink.

The heavy weight of the NinjaCU is not particularly welcome, especially as the mounting is not really a bolt-through-the-board solution. There are still hooks at the end of clips responsible for holding the >1kg weight on the motherboard. This could be more secure. The clip for the AM2 solution is also not quite secure enough for our liking, given its reliance on the two small plastic nubs on the stock retention bracket that come on AM2 motherboards. For this massive Fifth Anniversary model, Scythe really should have gone to a full through-the-motherboard spring-loaded-bolt system used successfully in the past by Swiftech and Alpha, and currently by Thermalright, as shown in the photo below.


Thermalright Ultra-120 and U120 Extreme use secure, high tension, spring-loaded bolt-through mounting.

Conclusion #2: The differences between the three Ninjas are most likely caused by the variances in tension or pressure applied between the heatsink base and the CPU by each mounting clip. There seems little doubt that the original Ninja's mounting setup applies the highest pressure. This might actually be felt during installation, although we have no way of knowing for sure, as measuring the pressure when the heatsink is engaged would be quite a challenge. The CU's mounting system seems pretty close, but the small size of the tabbed hooks is not reassuring, and it cannot provide as much pressure; the thermal results speak for themselves. The pushpins of the 775 socket mounting for the current standard Ninja are quite poor in comparison. The move by many SPCR forum members to use Thermalright's LGA775 spring-loaded bolt-thru kit is an excellent way to improve the mounting setup of the Ninja with Intel CPUs. Surely, the tighter, more secure installation would improve cooling by at least a few degrees.

AN ASIDE ABOUT NINJA PLUS REV B.

It seemed possible that our new sample of the Ninja Plus Rev. B is a poorer performer than is typical . It's almost unthinkable that a minor variant of one of SPCR's most respected low-airflow coolers could produce a whopping 24°C rise over ambient with our reference 120mm fan at full speed. That's mediocre performance for such a celebrated product. Out of curiosity, we also tried mounting it with its own standard socket 775 pushpins. The results were a touch worse.

Would a better fitting clip give better results?

A Thermalright LGA775 bolt-thru kit was not handy, but we did have the backplate and spring-loaded bolts from the Ultra 120 heatsink package. The Ninja's 640g mass is close to the U-120's 745g, so the tension of the springs should be about right. We ripped off the plastic pushpins from the 775 mounting bracket on the Ninja Plus Rev. B, and install the heatsink using the Thermalright backplate and spring-loaded bolts. The procedure was fairly straightforward: Just engage the threads on one bolt, then do the same on the opposite corner, and finally engage the remaining bolts. Tighten the screws in opposite corner pairs, going back and forth in turn. The notch at the corners of each fin on the Ninja makes it easy to apply a long screwdriver. Simply turn the screw until it stops, then each spring is compressed to the same degree.


The performance improvement was satisfying to record:

Scythe Ninja Plus Rev B: °C Rise w/ various clips
Reference 120mm Fan
12V
9V
7V
5V
Stock 775 pushpins
25
27
30
32
478 clip & NinjaCU hardware
24
25
28
31
Thermalright bolt-thru kit
21
22
24
26

Unfortunately, no 775 socket pushpin mounting clip is supplied with the NinjaCU, so this mod can only be used with the Ninja Plus Rev. B. Perhaps Scythe will get the hint.



Previous 1 2 3 4 5 6 7 Next

Cooling - Article Index
Help support this site, buy from one of our affiliate retailers!
Search: